Aaron Marketing Skills: Claude Marketing Skills Library
/aaron-marketing-skills
Aaron Marketing Skills is a 120-skill Claude Skills library plus 8 slash commands that give an AI agent scored, structured workflows across seven marketing disciplines, from SEO/GEO to influencer outreach. Reach for it if you want auditor-graded marketing work inside Claude Code without a SaaS subscription. Skip it if you need proof these skills already work on real campaigns — the README's own validation cohort for that hasn't run yet.
What are Aaron Marketing Skills?
Aaron Marketing Skills is a GitHub repository of Claude Skills and slash commands that equips a chat agent with structured, scored marketing workflows across seven disciplines — brand narrative, SEO/GEO, social, email, paid ads, influencer, and product launch. Every skill follows the same seven-section contract, and a shared protocol layer ties registries and memory together, so the 120 skills act as one system rather than disconnected prompts.
Key Capabilities and Principles
- ✓120 skills total: seven disciplines each contribute 16 skills (112) plus an 8-skill protocol layer, and every skill self-declares a `discipline` and `phase` in its metadata for routing.
- ✓Eight scoring frameworks — TALE, CORE-EEAT, CITE, STAR, ROAS, SEND, RAMP, ECHO — each enforced by an auditor-class skill that returns a machine-checkable SHIP/FIX/BLOCK-style verdict instead of a subjective read.
- ✓A shared veto policy: one verified veto caps a score at 59; two or more verified vetoes force a BLOCK verdict with no final score, and missing evidence resolves to 'Unknown' rather than an automatic fail.
- ✓Seven append-only registry streams (entity, creator, offer/claims, consent, launch, channel, narrative), each with a sole-writer skill — other skills submit proposal events through `registry-events.py` instead of writing directly.
- ✓Temperature-tiered memory: a HOT cache capped at 80 lines or 25 KB, whichever trips first, auto-loads every session, with per-skill WARM state underneath it.
- ✓Three capability profiles — Lite, Pro, Governed — gate connectors, saved audits, and registry writes, while the public command surface (`/auto` plus seven discipline commands) stays identical across all three.
- ✓Keyless by default: every skill runs on data you paste in or pull from free sources, and the paid-ads skills score off your own manual ad-account export rather than requiring a paid ad API key.
- ✓Installs on 70+ agent hosts via `npx skills add`, plus a dedicated Claude Code plugin path and a downloadable Portable Lite archive for Agent Plugins v1 clients.
Who Can Benefit?
Aaron Marketing Skills suits marketers or founders already running Claude Code (or another Agent Skills-compatible host) who want scored, repeatable workflows instead of ad hoc prompting — someone auditing SEO content, screening influencer fit, or building a launch checklist who's willing to paste in their own exports and read an auditor verdict. It's less useful if you want a managed platform with paid connectors built in: the Lite default works from pasted or freely-sourced data only, and the Pro/Governed profiles that add connectors and persistent audit trails require opting up deliberately.
Getting Started
For Claude Code: run `/plugin marketplace add aaron-he-zhu/aaron-marketing-skills`, then `/plugin install aaron-marketing@aaron` — the marketplace-add step only registers the catalog, so the install step is what actually turns the skills on. For the 70+ other Agent Skills-compatible hosts the README links (Codex, Cursor, OpenCode, Antigravity, Gemini CLI, Copilot CLI, OpenClaw, Hermes, and more), run `npx skills add aaron-he-zhu/aaron-marketing-skills`, or pull a single skill with `npx skills add aaron-he-zhu/aaron-marketing-skills -s keyword-research`. Agent Plugins v1 / Portable Lite clients can instead download the `aaron-marketing-skills-19.2.0-agent-plugin-v1-lite.tar.gz` archive from the v19.2.0 release and unpack it. Users on SkillHub.cn can install through `skillhub install <frontmatter-slug>`. Any host can fall back to a plain `git clone https://github.com/aaron-he-zhu/aaron-marketing-skills`. Note that installing the plugin adds nothing to your `/mcp` list — the MCP catalog in `docs/mcp-catalog.json` is kept outside the auto-registered path as a copy-paste reference.
Running Marketing Skills
If your host supports automatic skill routing, just describe the goal in plain language — the README's own examples include researching keywords for a SaaS product or finding TikTok creators for a skincare launch. Otherwise call `/aaron-marketing:auto` to let it infer intent and run the smallest workflow that fits, or go straight to a discipline entrypoint like `/aaron-marketing:seo-geo https://example.com/blog/my-article --phase tune` or `/aaron-marketing:ad` for paid-campaign work. Each of the seven disciplines runs its own four-phase loop — SEO/GEO goes survey → implement → tune → evaluate, for instance — ending in an auditor skill (`content-quality-auditor` for SEO/GEO) that scores the work before you ship it. A fresh install runs the Lite capability profile by default, even with the full plugin installed; Pro and Governed, which add connectors and persistent audit writes, are opt-in through a `--profile` flag, an `AARON_MARKETING_PROFILE` env var, or a project config file.
Strengths
- ✓One shared seven-section contract across all 120 skills means you learn the skill shape once — trigger, quick start, skill contract, handoff summary, data sources, instructions, next-best-skill — and it applies everywhere.
- ✓Auditor gates return a structured verdict (SHIP/FIX/BLOCK, or an outright BLOCK on two or more vetoes) instead of a vague quality score, so you know exactly what failed and why.
- ✓Keyless-by-default design lets you run the SEO/GEO or influencer-scoring flow from pasted data with no API keys or paid tools to provision first.
- ✓Apache-2.0 licensing gives you a permissive, business-friendly license with no copyleft terms to work around.
- ✓The registries' sole-writer rule plus append-only event streams give you an auditable trail of what changed — entity facts, creator terms, consent state — instead of one mutable state blob.
Current Limitations
- △The README calls this an 'engineering-validated' release — it clears CI, package-reproducibility checks, and simulated-fixture model runs — but draws a clear line under that: none of it counts as evidence the skills work on real client projects, and closing that gap is explicitly left to a future cohort of 14 pilots, 70 paired Lite/Governed comparisons, and 28 shadow projects that hasn't run yet.
- △This is one developer's project (aaron-he-zhu) rather than an organization-backed tool, and the README gives no adoption numbers beyond its own GitHub star count, so how many teams run it in production isn't documented.
- △Paid-ads scoring reads your own manually exported account data rather than a live ad API, so there's no automatic refresh — you re-export and re-run the skill whenever you want an updated read.
- △Agent Plugins v1 clients can't just clone the repo root; they need the separate Portable Lite release archive, which adds a step beyond a plain `git clone`.
- △Governed-profile features (registry writes, verified run evidence) ship in the current release, but the README treats Governed-by-default as a promotion still gated on the pilot cohort above — so anyone running Governed as their standard mode today is ahead of what the project itself calls proven.
- △The README verifies 120/120 skills installable across its listed hosts as of 2026-07, but exactly how each of the 70+ hosts degrades outside the full plugin is documented in a separate file (`docs/agent-compatibility.md`), not spelled out in the main README.
Alternative AI Marketing Tools
Frequently Asked Questions
Aaron Marketing Skills is released under the Apache-2.0 license, as listed on its GitHub repository.
Aaron Marketing Skills has a dedicated Claude Code plugin path, and the README lists 70+ other Agent Skills-compatible hosts — including Codex, Cursor, OpenCode, Antigravity, Gemini CLI, Copilot CLI, OpenClaw, and Hermes — installable via `npx skills add`.
Aaron Marketing Skills does not require external APIs by default — it runs keyless at Tier 1 on data you paste in or pull from free, first-party sources, with paid connectors as an opt-in convenience rather than a precondition, per the README.
Aaron Marketing Skills' README is translated into German, Spanish, French, Italian, Japanese, Korean, Portuguese, Simplified Chinese, and Traditional Chinese alongside the original English.
Aaron Marketing Skills routes each discipline's output through one of eight scoring frameworks — TALE, CORE-EEAT, CITE, STAR, ROAS, SEND, RAMP, or ECHO — via an auditor skill that returns a structured verdict, with a shared veto policy that caps or blocks scores when specific checks fail.
Aaron Marketing Skills is not yet validated on real-world projects, per its own README: the release clears internal CI and simulated-fixture checks, but real outcome evidence is pending a planned pilot-and-shadow-project cohort that hasn't launched.
The problem it solves
Most AI-agent 'marketing skill' add-ons are a folder of independent prompts — a keyword-research prompt here, an ad-copy prompt there — with no shared way to know if the output is any good or to hand context from one prompt to the next. Aaron Marketing Skills' README frames its reason for existing around that gap: one shared seven-section contract across all 120 skills, plus eight auditor gates that score the work instead of just generating it, so a skill's output carries a verdict and a defined next skill to hand off to.
Best use cases
- •Running an SEO/GEO content audit through `/aaron-marketing:seo-geo` and getting a CORE-EEAT/CITE-scored SHIP/FIX/BLOCK verdict from `content-quality-auditor` before you publish.
- •Screening influencer fit for a campaign — the STAR framework's `fit-scorer` produces a Suitability read, and `creator-content-auditor` gates creator content before it goes live.
- •Scoring your own paid-ad account exports against the ROAS framework via `ad-account-auditor` before scaling budget, without connecting a live ad API.
- •Drafting and gating brand-narrative claims through the TALE framework's trace → architect → land → evaluate loop, so narrative work gets separate truth, system, and effectiveness reads instead of one blended score.
- •Coordinating a product launch with the RAMP framework and `launch-readiness-auditor`, using `launch-registry` to track stage, dates, and submissions in one append-only stream.
Related repositories
Is aaron-marketing-skills worth your time?
ChatGPT, Claude and Perplexity can all read this page. Ask one of them what it makes of aaron-marketing-skills.